AI and the Path Toward Making Sense of Everything

As the industry grapples with the energy trilemma of reliability, affordability and sustainability, managing production at oil and gas fields has never been more important.

But optimally managing that production has not been easy because of siloed data. Adding to the complexity, each reservoir is different. Each development is distinctive. And technology selection varies widely.

Recent advancements in digital technology, such as artificial intelligence (AI), machine learning (ML) and cloud computing have opened up new avenues to create scalable, automated and data-driven oil and gas production. 

Leucipa, created by Baker Hughes and Amazon Web Services (AWS), is named after the Greek philosopher Leucippus, who observed: “Nothing occurs at random, but everything for a reason and by necessity.”

But understanding how everything in, for example, an oil field, fits together is no small task.

Leucipa combines AWS cloud capabilities with Baker Hughes’ oil and gas expertise in a product that the companies say will help operators optimize production of oil and gas fields.

“Production is a complex domain,” Howard Gefen, general manager for energy and utilities at AWS, said.

The industry has done a good job of optimizing the individual parts of production, but due to siloed data and solutions, has not been able to “make sense of the interdependencies” of all the pieces that feed into optimizing a field’s production, he said.

“With digital solutions, with data and AI and ML tools, we can now sort of make sense of everything and provide data-driven insights and recommendations,” Gefen said.

‘Scattered approach’

James Brady, chief digital officer at Baker Hughes, said operators are looking for ways to use technology to reduce lifting and operating costs, often through opex optimization or extension.

“It needs to be technology that makes an impact and pays for itself,” he said. “You need to look at the technology and tools that allow energy companies to be able to produce and extract and make their production commitments, but with less people.”

He said production software has been “very scattered” in the past and there’s not been a single company that covers all the bases. Some perform tasks competently: well or reservoir forecasting, history matching or bespoke field solutions, he said.

“You just see a real scattered approach, and it tends to be along the technology axis. So companies divide different parts of the problems. Nobody attacks the whole thing,” he said. “Baker Hughes has some pieces, but we don't have it all, either.”

In the industry, variation is the rule: particularly field to field and with many different pieces of equipment at work.

“That's part of what makes production software hard,” Brady said. “It’s not a product play with a standard template. The other thing that makes production hard is every field is different with respect to its infrastructure. You kind of have to go in there with a light touch, be able to connect and automate and then quickly be able to get in there and do something with data quickly that's relevant to that field.”

Baker Hughes wanted to listen to the market and put together technology that addressed the unmet production software needs, so the company commissioned McKinsey & Co. to carry out a market study on what companies need and want from production software — and which needs were unmet.

“What's the number one production software tool? I think probably most of us that have been around know that the dominant market player is [Microsoft] Excel spreadsheets,” Brady said.

Getting to market quickly meant collaborating, he said.

“We're not a cloud company —AWS is, and one with a strong technology vision,” an understanding of solutions for energy and the ability to help apply patterns they have seen elsewhere, Brady said. AWS knows how to build scalable solutions and has experience building AI and ML frameworks in different domains, he added.

“In some sense, they're an enabler, but they're also a technology partner to help guide us to the next step of this journey,” he said.

Gefen said Leucipa leverages advanced analytics and AI and ML tools.

Leucipa is “an opportunity for us to leverage what's available now in terms of cloud architecture and open architecture to create a platform that brings all these silos and bits of data and these interdependencies together,” he said.

The result is a solution that can scale and automate production operations, he said.

After the early adopters bring their data into Leucipa, “they'll be able to learn something about what that data is telling them that they should do. The initial focus is their wells, how they can be optimized on a single-well basis and eventually a set of wells together,” Brady said. “One of the things that cloud really enables us to do is this ability to look at a lot of scenarios.”

Sophisticated modeling is one of the more interesting scenarios made possible through automation—whether on increasing production or holding it steady while decreasing carbon emissions, Brady said.

“It's a different type of constraint-based modeling than we did before,” he said. “I think that will require the cloud to really explore the solution space that helps us honor those two parts of the trilemma.”

Baker Hughes and AWS announced in February their Leucipa solution and Pan American as the first company to publicly commit to working with Leucipa. Since then, nearly a dozen companies have signed on.

The initial release is still under development, but the expectation is that the early adopters will get their first taste of Leucipa around mid-year, Brady said.